The purpose of the study is to develop and validate a tool for reliable individualized prognostication of Stage III melanoma patients for use in the clinical setting. Cutaneous melanoma is the sixth most common cancer in the United States, and its incidence rate is increasing faster than any other cancer. Nearly 69,000 new cases are expected be diagnosed in this country in 2010. While thin melanomas are typically cured with excision alone, thicker melanomas have a greater tendency to metastasize to the regional lymph nodes. A diagnosis of Stage III melanoma is made if there is spread to the regional lymph nodes. Unfortunately, there is marked diversity in the natural history of Stage III melanoma, and outcomes within this group are extremely heterogeneous, with 5- year survival rates ranging from 23% to 87%. Similarly, treatment options range from intensive forms of systemic therapy to observation. Understanding patients' differences in clinical outcome is critical not only for calibrating therapeutic intensity to metastatic risk but also in the design and analysis of clinical trials. There is a real void of reliable prognostic tools for Stage III melanoma. Based on novel machine learning approaches, the purpose of this study will be to develop and validate a reliable and individualized tool for prognostication of Stage III melanoma patients that can be used in the clinical setting.